Incident 61: Overfit Kaggle Models Discouraged Data Science Competitors

Description: In the “The Nature Conservancy Fisheries Monitoring” competition on the data science competition website Kaggle, a number of competitors overfit their image classifier models to a poorly representative validation data set.
Alleged: Individual Kaggle Competitors developed and deployed an AI system, which harmed Individual Kaggle Competitors.

Suggested citation format

Yampolskiy, Roman. (2017-05-01) Incident Number 61. in McGregor, S. (ed.) Artificial Intelligence Incident Database. Responsible AI Collaborative.

Incident Stats

Incident ID
61
Report Count
1
Incident Date
2017-05-01
Editors
Sean McGregor

Tools

New ReportNew ReportDiscoverDiscover

CSET Taxonomy Classifications

Taxonomy Details

Full Description

On the data science competition website Kaggle, a number of competitors in the “The Nature Conservancy Fisheries Monitoring” competition overfit their image classifier models to a poorly representative validation data set. This resulted in intermediate competition rankings that were misleading and discouraged other data scientists from competing. Outside of the competition environment it would not have been clear that this error had taken place.

Short Description

In the “The Nature Conservancy Fisheries Monitoring” competition on the data science competition website Kaggle, a number of competitors overfit their image classifier models to a poorly representative validation data set.

Severity

Negligible

Harm Distribution Basis

Religion

AI System Description

Image classifer models designed by individual competitors on Kaggle.

System Developer

Individual Kaggle Competitors

Sector of Deployment

Public administration and defence

Relevant AI functions

Perception

AI Techniques

supervised learning, machine learning, DNN, VGG, open-source

AI Applications

Feature detection, Image classification, Decision support

Location

Global

Named Entities

Kaggle, The Nature Conservancy

Technology Purveyor

Kaggle Competitors

Beginning Date

2016-11-14T08:00:00.000Z

Ending Date

2017-04-12T07:00:00.000Z

Near Miss

Near miss

Intent

Accident

Lives Lost

No

Data Inputs

Images captured on fishing boats

Incidents Reports

What I’ve learned from Kaggle’s fisheries competition

Gidi Shperber Blocked Unblock Follow Following May 1, 2017

TLDR:

Me and my Kaggle partner, have recently participated in “The Nature Conservancy Fisheries Monitoring” (hereby: “fisheries”) Kaggle competition. This competition had some very interesting results, most notably, the benchmark result got ahead of most competitors. I believe there are a few very important lessons we can learn form as data scientists. Bellow you can read about the competition, and some of the things we have learned from it.

The Competition

So recently you’ve might have heard some buzz about Kaggle’s fisheries competition, and you’ve might have been wondering what all the fuss is about.

In this competition your goal (as in many Kaggle competitions) was to classify an object, in this case - determine the species of a fish.

why would you like to do this? Well, to make a long story short, ships are currently equipped with automatic cameras, to make sure they are catching only legit fish like tuna, and not protected animals like sharks. The picture are later manually deciphered

However, the fish are not presented to you in a clear profile pic, but it is randomly tossed around some boat. Here are some example images:

there are also images with more than one fish, or images with no fish at all.

So this is not the standard “recognize the flower/leaf/face” competition, but a more challenging one.

And this is why I love Kaggle: unlike many machine learning articles, which use the easiest available data-set (mnist, cifar-10, imagenet to name a few) Kaggle presents you with real life problems, which are always dirty, mislabeled, blurred etc.

As in some other competitions, this was a two stage competition, in which a small part of the test data (1K images) is available, while the majority of the test set (another 12K images) is released 1 week before the competition deadline. This method may prevent some cheating, but has a lot of issues, and may discourage many competitors: only around 300 competitors of the total 2300 starters made a submission at the second stage.

Our way through the competition

From a quick look, it is easy to see that the required approach here is doing two steps: first detecting the fish, and then classifying it.

However, during the competition, something strange came up in the competition forums: if you run a pre-trained VGG network on the training set (as very clearly described here and more thoroughly discussed in this great course), you can reach accuracy of around 99%. This is strange, because the VGG was trained on image-net data, which is not very similar to the images above. Submitting these results to Kaggle showed results which were not bad at all, but much lower than 99%.

What I’ve learned from Kaggle’s fisheries competition

Similar Incidents

By textual similarity

Did our AI mess up? Flag the unrelated incidents